DeepSeek's AI Assistant has risen to prominence, becoming the most-downloaded free app on Apple's App Store in the United States, surpassing even OpenAI's popular ChatGPT. Founded in 2023 by Liang Wenfeng, a prominent figure in the AI industry and co-founder of the quantitative hedge fund High-Flyer, DeepSeek is making significant waves in the technology sector. The company has gained attention for its cutting-edge R1 model, which rivals OpenAI's o1 model both in performance and innovation.
DeepSeek's commitment to advancing artificial intelligence is evident in its substantial hardware expenditures, estimated to be "well higher than $500M over the company history" according to SemiAnalysis. The company's R1 model, praised for its capabilities, is an open-source platform that invites AI developers worldwide to utilize and enhance it. This openness has not only fueled its popularity but also garnered accolades from industry leaders, including OpenAI CEO Sam Altman.
"This is a reminder of the level of competition and the need for democratic AI to win" – Sam Altman
The R1 model's emergence as a formidable competitor to established models underscores DeepSeek's strategic investments in research and development. Despite training costs amounting to almost $5.6 million, these expenses are significantly lower than those incurred by its competitors. The company's ability to achieve high performance with relatively modest spending is seen as a testament to its innovative approach.
"To be clear DeepSeek is unique in that they achieved this level of cost and capabilities first" – SemiAnalysis
DeepSeek's focus on artificial general intelligence (AGI), a branch of AI that matches or surpasses human intellect across various tasks, has positioned it at the forefront of technological advancement. Released in April 2023, the R1 model has captured public interest, with its popularity surging earlier this month. However, despite its impressive capabilities, the model does face limitations, particularly when addressing topics related to Chinese leader Xi Jinping's policies.
The company's strategic direction and financial support come entirely from High-Flyer, providing a stable foundation for continued innovation. Generating "synthetic data" for training purposes requires a "considerable amount of compute," contributing to the company's significant R&D costs and total cost of ownership.
"The level of interest in reasoning, the level of interest in open source" – Sam Altman
DeepSeek's achievement in reaching such levels of performance and capability with relatively lower costs sets a new benchmark within the industry. Its hardware investments and open-source approach indicate a forward-thinking strategy aimed at fostering collaboration and innovation.